Where is Interconnect Technology Heading?

Written by Deepak Chandra Sekar

With 16nm chips moving to production this year, companies are actively developing the 10nm and 7nm technology nodes. These generations are interconnect heavy with more than 50% of their cost due to Back-End-Of-Line (BEOL) wiring levels, and designs that are dominated by interconnect delay. Qualcomm’s VP of Technology, Geoffrey Yeap, in fact, called BEOL performance/area/cost the foremost issue for 10nm and 7nm nodes. Engineers are taking several paths to work around this trend. These paths were discussed at the recent joint conference of the IEEE International Interconnect Technology Conference (IITC) and the Advanced Metallization Conference (AMC).

Having attended the conference, I’ve outlined the main technical trends here:

• Copper barrier/cap/liner optimization a focus area – In wiring levels, you have copper, the low k dielectric and a whole bunch of materials that help interface between them. These include barrier materials, liners and cap layers for copper. This year’s conference had a plethora of work on the subject, including Co and Mn-based materials for these layers, evaluation of new process techniques such as ALD as well as studies on thinning these layers. Up to three orders of magnitude improvement in electromigration time-to-failure, as well as significant performance improvements were demonstrated.
• Low k dielectric improvements slowing down – There was a general sense that dielectric constant improvements were slowing down.
• Through-silicon via (TSV) technology – GLOBALFOUNDRIES provided updates on their 2.5D/3D manufacturing readiness and also presented a neat technique to eliminate “keep-out zones”. The technology is ramping up for production at many companies. I was delighted to see this technology that I started working on 12 years back finally move to the production-phase!
• BEOL memory – Intel gave updates on the stacked capacitors they were integrating for their embedded DRAM process. Panasonic described the Resistive RAM microcontroller they are selling as well, which provides significant power and cost benefits over embedded flash microcontrollers. Expect such BEOL memories to become commonplace in the future as memory demands grow from the Internet of Things and Big Data applications.
• Life beyond copper – Copper resistivity rises exponentially at scaled nodes due to surface scattering, grain boundary scattering and diffusion barrier thickness. This has been motivating research beyond copper solutions throughout the industry. There were papers demonstrating carbon-based interconnects, such as graphene and nanotubes that could provide improved performance and reliability compared to copper. However, several process integration challenges need to be resolved for these solutions to become practically viable. Metal alternatives to copper are also being studied.
• Interconnect-aware design – For long interconnects, due to the use of repeaters, performance and power, depend as much on the transistors as the interconnects. Several speakers discussed optimization of repeaters in multi-level interconnect stacks. As well, there were hallway conversations surrounding methods of applying Network-on-chips to tackle interconnect issues.

The conference was very successful this year with superb attendance, a top-class program and some fun events in the evenings. As General Chair of this year’s conference, it was a relief to have a successful event! I’d like to thank the members of the IITC/AMC committees – it was a lot of fun working together to organize the conference.

One of my favorite parts of the conference was the Best Student Paper Award. I received the Best Student Paper award at IITC myself many years back. This was one of the more memorable events in my grad school days so I like to keep tabs on this award, and, in fact, Rambus was proud to sponsor this award. We’re happy to share that The Bakir Group from Georgia Tech received the award this year for their work on liquid cooling 3D stacked chips.

Thanks again to all the organizers and participants of this year’s conference – see you next year!

The Best Student Paper Award, which was sponsored by Rambus, was given to the Bakir Group at Georgia Tech. I’m on the left, giving the award to Prof. Muhannad Bakir (right).